Seam tracking in electron beam welding - a new detector

TWI Bulletin, March/April 1988

Allan Sanderson, PhD, CPhys, FInstP, CEng, FIM, AMCST, FWeldl is Head of the Electron Beam Department.

Colin Ribton, BSc (Hons) is a Research Physicist in the Electron Beam Department.

Real time seam tracking in electron beam welding has obvious advantages for automation and can prevent operator fatigue in high production rate welding. A new detector system is described which was developed at The Welding Institute and uses back-scattered electrons to allow real time seam tracking, even at high beam powers.

The detection of back-scattered electrons as a means for imaging and locating surface features dates back to the development of scanning electron probe micro-analysers and scanning electron microscopes. With the low current beams used, enormous magnification of surface details is readily achieved. The application of similar techniques for detecting joint line features for real time seam tracking in electron beam welding, but using the full welding beam, has been a tantalising possibility for a long time. Unfortunately, particularly with thick section components, the high power levels are such that the detected signal is subject to severe interference - largely because of the copious generation of positive ions and charged spatter. Recent work at The Welding Institute has concentrated on the development of real time seam tracking using a new detector which greatly improves signal integrity, especially at high operating powers.

Electron beam welding is a process which lends itself to computer control, since all the process variables can be readily converted to digital parameters. For example, in contrast to arc welding, fusion zone shape and penetration are much less dependent on workpiece geometry and heat source/material interactions. However, although the high depth-to-width ratio of an electron beam fusion zone can be used to great advantage, this characteristic of the process imposes stringent beam/joint alignment requirements.

In the vast majority of electron beam machines, alignment is solely under the visual control of the operator. Although for small or simple components, where the joint line is a short linear or circular seam, this task is trivial, with large objects seams can deviate from an idealised geometry because of machining tolerances. In addition, where the total heat input relative to the component and jigging mass is large, the joint line may even distort during welding. These problems preclude the use of 'teach and playback' methods of seam tracking, therefore there is a need for real time control. In addition, where component throughput is high, such as in the fabrication of rotor blade assemblies for aerospace components, there is a practical advantage to be gained by automatic seam tracking; that of alleviating operator fatigue.

Unexpected deviations of the beam can also occur during welding because of residual magnetism in the workpiece or jigging. Similarly, in welding certain dissimilar metals, thermo-electric currents can give rise to severe magnetic fields. In both these cases, automatic seam tracking is used to correct beam/joint misalignment at the work surface. However, for thick section welding, angular displacement can give rise to 'missed joint' defects towards the base of the weld. It is therefore recognised that extraneous magnetic fields should be eliminated for seam tracking to be fully effective on thick section components.

Seam tracking methods

Mechanical

Attempts have been made to use a mechanically mounted probe ^[1] which runs along a seam. This means that the centre of the probe has to be aligned with the beam. Such mechanical probes lack versatility in that each one can only be applied to a limited range of component geometries. Also such pre-aligned probes take no account of magnetic deflection effects.

Optical

Automatic seam tracking can be achieved by light optics, using a high intensity lamp or laser beam directed at the joint line. Light sensitive detectors placed above the workpiece are used to sense the joint. Such optical methods could be satisfactory at modest beam power levels, provided that the stray light from the molten pool can be eliminated. However, at higher powers, coating of the optical elements and light source with metal vapour is a substantial problem. ^[2] Moreover, the sensor elements have to be protected from damage by the intense heat and X-ray emission from the molten pool zone.

Electrical

The change in 'through-current' produced when the electron beam is swept across the joint line can be used to determine the beam position relative to the joint. ^[3] This system is only applicable to relatively thin materials and where there is access for positioning a 'through current' detector behind the joint.

X-ray emission

In a technique similar to the above, the joint can be located by the change in X-ray emission as the beam is scanned across the joint line. For best results X-ray detectors are placed close to the workpiece surface, but set laterally some distance away from the molten pool. Several papers have been published on this technique, but to date it has found little industrial application. ^[4]

Back-scattered electrons

When a charged particle beam consisting of particles such as electrons impinges on the surface of a material, a fraction of the particles is backscattered. The size of the fraction is dependent on the angle of incidence of the beam on the workpiece, the atomic number of the workpiece elemental constituents, and the energy of the incident particle beam. ^[5]

In the case of an impinging electron beam, the energy of the back-scattered electrons can be described as a continuous spectrum varying from a fraction of an electron volt to the full incident energy of the electrons in the primary beam.

The proportion of back-scattered electrons is also dependent on the workpiece topography. If the electrons impinge on any part of the workpiece which is not substantially normal with respect to the incident beam, then the back-scattered electrons are generally directed away from the source of the incident beam.

The back-scattered electron seam tracking method is based on the principle that there is a reduction in the quantity of electrons reflected by the workpiece as the beam is swept across the joint. By timing the delay from the start of a pre-determined sweep to the point at which electron loss occurs in the backscattered signal, the relative position of a joint line with respect to the welding beam null position can be computed and appropriate servo corrections made through either traverse or beam deflection controls.

Seam tracking by back-scattered electron emission

In scanning electron microscopes ^[6] the low current primary electron beam is scanned in a raster mode over the workpiece surface. The back-scattered electron current (which can be collected by a simple detector placed above the workpiece surface) is then used to modulate the brightness of a synchronised raster on a TV monitor to produce high contrast and high magnification electron images. The corresponding seam tracking method makes use of a simplified version of this technique whereby the welding beam is periodically deflected out of the molten pool and scanned across the joint line ahead of the weld ( Fig.1). The back-scattered electrons are generally detected by a simple fixed plate surrounding the beam. ^[7] Where the beam power and power density are relatively low, such a simple detector can provide a clear indication of the joint line position. However, as power levels and power densities are raised, the quantity of positive ions generated by the molten pool increases. These positively charged particles back-stream towards the detector plate and can cause degradation of the collector signal. Attempts have been made to control the flow of positive ions by inserting an intervening grid which is biased negative ( Fig.2). ^[8] Unfortunately it is necessary to have either a high biasing voltage or a relatively fine mesh size to attract the positive ions. In the presence of ionised gases and metal vapour it is difficult to avoid breakdown of the high voltage grid. On the other hand, a fine mesh tends to become rapidly clogged by weld metal spatter and vapour emitted by the molten pool.

The intense and erratic emission of positive ions of varying energy from the molten pool has hindered the use of the back-scattered electron technique. Particularly, difficulties have been experienced in discriminating between the joint signal pulse and other molten pool emissions. One particular attempt to overcome this problem has been described by The Paton Institute, in which an auxiliary low power electron gun and deflection system is run concurrently with the main welding beam. This allows the detector to be positioned such that only the low power sweep is monitored, thus avoiding interference from the molten pool emission. ^[9] The expense and extra complication of using an auxiliary gun with independent deflection and focusing coils makes this system less desirable, although to date it is the only tracking method that has been claimed to work for beam powers up to 100kW.

Performance with conventional detector

All tests were carried out in the flat position (beam axis vertical) with an accelerating voltage of 150kV ( Fig.2). A 1k Ω grounding resistor, connected to an oscilloscope, was used to monitor the collected backscattered current.

Effect of pool melting

It was suspected that the presence of the molten pool would affect signal appearance and integrity. Therefore tests were carried out with a static beam at a given power on a stationary mild steel testpiece, but with three different power densities resulting from different degrees of focus. For a greatly defocused beam (which caused heating but no visible melting of the workpiece) a steady electron emission was obtained as indicated by the oscilloscope trace in Fig.3a. However, when the workpiece started to melt ( Fig.3b) the signal fluctuated with a fairly constant amplitude of some 1 kHz frequency. With substantial melting and vaporisation, the signal amplitude and frequency became much more erratic, Fig.3c.

Some of the fluctuations in collector current can be explained by positive ion neutralisation of the backscattered electron stream. In addition, violent boiling of the liquid metal caused erratic angular changes with respect to the impinging beam with a corresponding reduction in collector current. Furthermore, the formation of a keyhole acts as an efficient particle trap for the back-scattered current. Thus the characteristic rapid opening and closing of such a keyhole would cause corresponding gross fluctuations in electron back-scatter as in Fig.3c.

Effect of extraneous emission

For a more realistic appraisal, tests were made with a 40mA (6kW) beam scanned rapidly across a joint groove 1mm wide and 10mm deep.

Figure 4a shows the back-scatter signal during a scan. At point 'A' the beam is deflected out of the molten pool with a corresponding increase in the total back-scatter current level. At point 'B' the transverse deflection is maximum value, and the beam at its furthest from the molten pool. Then, as the beam moves across the artificial groove, a loss of electrons occurs producing the current drop at 'C'. The beam was then swept to the opposite extremity and back into the molten pool (D, E), as before.

At this welding current it appears that there is little interference from the molten pool during the scan, but Fig.4b shows that erratic emissions are occurring from the molten pool, which are confirmed in Fig.4c for a succession of scan cycles. It is clearly difficult to discriminate between the true signal, detected during a scan, and the molten pool interference, except by cross reference to the transverse deflection coil current. The average level of the signal during the scan time also changes somewhat from one cycle to another.

Effect of beam current amplitude on collected signal

Oscilloscope traces were obtained for the same 1mm wide by 10mm deep joint preparation, for beam currents up to 200mA (30kW). The beam was focused at the surface of the workpiece and no melting of the block surface occurred during the scan cycle even at the maximum current used. At relatively low currents, e.g. 20mA, a sharp pulse corresponding to the joint feature is obtained ( Fig.5a) although there were still large fluctuations in signal level outside the scan period. At 60mA a similarly clear indication was observed ( Fig.5b), and at the appropriate sensitivity the molten pool emissions were relatively lower than the scan signal. This trend continued as the beam current was increased further to 120mA ( Fig.5c) but it was also noticed that the back-scatter signal itself had been degraded. These effects were also observed to a greater extent at 200mA, Fig.5d. Furthermore, the average DC level of the signal during the scan time was noted to fluctuate from one scan to another for all of the beam currents examined. From these results it was decided that the emission of positive ions was affecting the back-scattered electron signal.

Performance with new detector

Design principles

An alternative design of detector plate was developed (UK and foreign patents applied for) to eliminate the effects of the positive ion contribution to total collector current in which the backscattered electrons are reflected onto a collector which is otherwise shielded from direct line of sight of the molten pool. This radically different design was based on the fact that high energy electrons can be reflected from metallic surfaces with little or no loss of energy, while positive ions can be easily absorbed once contact has been made with a metallic surface.

In the first version ( Fig.6) with flat reflecting surfaces, 50V negative potential was applied to the reflector and shield electrode enveloping the collector. In this arrangement, the particle stream in the annular solid angle Δ α impinges on the shield electrode and does not affect the collector plate C, whereas a proportion of the particle stream of annular solid angle Δ Θ reaches the reflector and arrives on to the hidden annular collector plate, C. Since the shield electrode and reflector are biased negative a high proportion of the relatively slow moving positive ions are absorbed (in addition it was anticipated that there would be a tendency for low voltage secondary electron and thermal electrons to be repelled, so that they would not enter the collector cavity).

To increase the zone of electron current collection (which is important when operating at low beam currents and long working distances), this novel design was further extended. Figures 7a and b show two hidden collectors, with specially shaped reflectors. These enable the device to be used over a wide range of working distances, e.g. from 100-1000mm. In addition, the use of two annular collector plates allows differential signal conditioning to be used to achieve noise reduction and greater immunity from signal level shifting. The use of two collectors is also beneficial in the case of electron image production since it permits better control of video signal contrast.

In this more advanced detector arrangement, the shaping and positioning of the reflecting surfaces AB and CD ( Fig.7a) are such that as much as practicable of back-scattered high energy electrons are collected with minimum obscuration by mechanical structure members. In addition, since the majority of the back-scattered electrons is concentrated in the plume of particles leaving the workpiece surface with near normal incidence, the total collector signal is larger than for the simple reflector design shown in Fig.6.

Effect of molten pool

The static beam experiments described above were repeated for the new design of collector system. Although the level ( Fig.8) of the signal (monitored across the 1k Ω resistor) was approximately a tenth of that obtained with the plain plate detector ( Fig.3a) it was observed that the average DC level of the signal did not appreciably change when surface melting just occurred ( Fig.8b). Moreover, when the beam was tightly focused on the surface to produce a rapidly fluctuating keyhole the average level of the signal only changed slightly, although there were major fluctuations in the signal ( Fig.8c). 

Thus the average signal level from a heated workpiece and from a molten pool, Fig.8b and c, is proportionally larger than previously (compare with Fig.3b and c) indicating that positive ions no longer reach the collector. However, it was found that the collected signal still displayed large amplitude fluctuations even with the hidden plate detector in use. In addition the signals for the simple plate detector and hidden detector were similar at low beam current.

Since the positive ions have been removed, the erratic signal for the focused beam must be caused by the molten metal movement with, at low beam currents, a relatively shallow keyhole which acts as an inefficient particle trap.

Effect of extraneous emission

The experiment of scanning a 1 x 10mm joint groove was repeated with a beam current of 40mA and a travel speed of 100 mm/min. With the new detector system there was little interference from the molten pool during the scan ( Fig.9). Moreover, during the welding time, Fig.9b and c, it was noticed that the interference from the molten pool was much reduced compared with that using the simple plate detector ( Fig.4b and c) because of the removal of back-streaming positive ions.

Effect of beam current

With a 20mA beam current a clear signal pulse was produced ( Fig.10) but outside of the scan time large fluctuations in signal occurred corresponding to molten pool emissions. As the current was increased to 60mA, Fig.10b, the relative amplitude of these fluctuations was reduced. A clear indication of joint line position was also obtained even at beam currents of 120mA, Fig.10c, and 200mA, Fig.10d. In addition, at these higher beam currents very little signal was collected outside of the scan time.

Effect of beam focus

Using the hidden collector detector, further tests determined the range of focal position over which joint detection was possible for the 1mm wide by 10mm deep groove. This was carried out at 20mA beam current and the signal current monitored across the 1K Ω resistor. The signal obtained for surface focus is shown in Fig.11a. With a reduced lens coil current giving a focus approximately 40mm below the workpiece surface, the signal still gave a good indication of joint position although some broadening of the pulse occurred, Fig.11b Similarly, with increased lens coil current (focus 40mm above the work) a reasonable signal was obtained, Fig.11c.

However, with focus positions approximately 80mm above and below the work, the pulse was substantially broadened, Fig.11d and e, such that the joint position was less clearly defined.

Further developments

As mentioned earlier, stray magnetic fields can cause 'missed joint' defects even with a seam tracking system based on back-scattered electrons. This follows since the tracking system is only able to place the beam on the joint line, and takes no account of the beam approach angle. For this reason in a further development a combined magnetic shield and detector arrangement has been constructed using the hidden collector principle ( Fig.12).

Subsequent to these developments further improvements have been achieved in signal processing to eliminate fluctuations in signal level caused by extraneous molten pool emissions, erroneous pulses produced by external signal interference and miscellaneous joint line features such as weld tacks and spatter.

Based on the work described, an industrial prototype real time seam tracking system has been designed and manufactured which is now being used in conjunction with a 100kW gun column for numerous applications. An example of the effectiveness of the system is indicated by the curvilinear specimen shown in Fig.13. This demonstration specimen comprised two 75mm deep, 9.5mm thick mild steel bars with a 0.8mm wide, 10mm deep mild steel shim which served as a joint feature. The specimen was successfully tracked at a beam power of 150kV x 64mA at a welding speed of 125 mm/min.

Acknowledgements

The work was funded jointly by Research Members of The Welding Institute and the Minerals and Metals Division of the UK Department of Trade and Industry.

References